pulse-shape analysis of pdm-qpsk modulation formats for
TRANSCRIPT
Pulse-Shape Analysis of PDM-QPSK Modulation Formats
for 100 and 200 Gb/s DWDM transmissions
Andrés Macho Ortiz · Paloma R.Horche
Abstract Advanced optical modulation format
polarization-division multiplexed quadrature phase shift
keying (PDM-QPSK) has become a key ingredient in the
design of 100 and 200-Gb/s dense wavelength-division
multiplexed (DWDM) networks. The performance of this
format varies according to the shape of the pulses
employed by the optical carrier: non-return to zero
(NRZ), return to zero (RZ) or carrier-suppressed return to
zero (CSRZ). In this paper we analyze the tolerance of
PDM-QPSK to linear and nonlinear optical impairments:
amplified spontaneous emission (ASE) noise, crosstalk,
distortion by optical filtering, chromatic dispersion (CD),
polarization mode dispersion (PMD) and fiber Kerr
nonlinearities. RZ formats with a low duty cycle value
reduce pulse-to-pulse interaction obtaining a higher
tolerance to CD, PMD and intrachannel nonlinearities.
Keywords Modulation formats · Quadrature Phase Shift
Keying · duty cycle · spectral efficiency (SE) ·
intrachannel nonlinearities
1 Introduction
Nowadays, communication networks are required to
provide an enormous data transport capacity to solve the
continuous increase of internet traffic. The amount of
traffic carried on backbone networks has been growing
exponentially over the past two decades. The required
network bandwidth increases between 40% and 60% per
year due to the rapid emergence of new communication
services: social networking, 4K video, data traffic of
smartphones and tablets, and cloud services. The
explosion of these services has led to the necessity of
implementing new technologies in optical transport
networks which increase their capacity [1,2,3].
Since the 90s, traffic demand in core optical networks
has been covered by WDM systems, which have been
able to scale up their capacity from 10-Mb/s per channel
to 100-Gb/s at present [4]. Migration from 40 to 100-Gb/s
has not been so immediate because it has required
additional transmission techniques that would allow this
bit rate compatibility with DWDM-50 GHz grid.
Polarization-division multiplexing (PDM) was developed
to increase spectral efficiency of QPSK modulation
format from 2 b/s/Hz to 4 b/s/Hz. In addition, the use of
coherent detection, digital signal processing (DSP) and
forward error correction (FEC) techniques, the standard
100 Gigabit Ethernet (GbE) became a reality at the end of
2010 and at the beginning of 2011 [2]. With the impulse
and development of Cloud Computing, the capacity
offered by DWDM networks could be short in the next
years, so optical communications should survey beyond
100-G, being the next step the standard 400 GbE,
estimated for the year 2016 [3,5,6].
A.Macho ∙ P.R.Horche ( ) Departamento de Tecnología Fotónica y Bioingeniería
Escuela Técnica Superior de Ingenieros de Telecomunicación
Universidad Politécnica de Madrid, Avda. Complutense nº30
28040 Madrid, Spain
A.Macho
e-mail: [email protected]
telephone: +34649994228
P.R.Horche
e-mail: [email protected]
telephone: +34913367306
fax: +34913367319
As previous step to the 400 GbE, the immediate
necessity is to migrate from 100-G to 200-G. To achieve
this bit rate is likely to continue initially with the parallel
transmissions approach: 5 parallel lanes of 40-Gb/s or 2
parallel lanes with quickly maturing 100-Gb/s technology,
[3]. But the tendency is to integrate 200-Gb/s in 50 GHz
grid through a single lane.
Therefore, high spectral-efficiency (S.E) PDM-M-
QAM formats have been proposed to increase the S.E
beyond 4 b/s/Hz. The main drawback of using dense
digital constellations is their high vulnerability to the
nonlinear effects of the optical fiber, increasing the
system complexity to achieve long-haul transmissions
with these formats [7]. The other disadvantage inherent to
the use of dense digital constellations is that it would be
necessary a higher resolution in the analogic-to-digital
(ADC) converters of the coherent receiver [8].
Due to the difficulties to obtain long-haul optical
transmissions with PDM-M-QAM formats [9], the
International Telecommunication Union Standardization
Sector (ITU-T) has enabled to leave the rigid grid of 50
GHz in 2012 with the proposal of flexible DWDM
networks [10]. In this way, it is possible to avoid the use
of PDM-M-QAM formats employing 200G-PDM-QPSK
signals with 100 GHz between optical carriers (Fig. 1).
For the integration in DWDM-50 GHz grid it is being
studied the Nyquist filtering approach (N-WDM), which
reduces the width of the main spectral lobe [11] at the
expense of a strong filtering distortion. This penalty could
be solved with Maximum-Likelihood Sequence Detection
(MLSD) [12,13]. However, the abrupt transition band of
Nyquist filtering is currently far from its implementation,
so the practical approach is to use these formats over the
grid of 100 GHz for 200-Gb/s transmissions.
The performance of PDM-(D)QPSK varies according
to the line code that is employed over the pulses of the
optical carrier. The question as to the ‘best’ optical
modulation format cannot be concluded until the pulse is
fully shaped: NRZ, RZ or CSRZ (Table 1). Each of the
options offers different tolerances to linear and nonlinear
optical impairments:
ASE noise
Crosstalk and optical filtering distortion
Chromatic dispersion
PMD
Fiber Kerr Nonlinearities
In this paper we will analyze the features of PDM-
(D)QPSK in 100 and 200-Gb/s transmissions heeding the
above physical impairments. The paper is organized as
follows. Section II presents the architecture of the
simulation scenario and its main parameters, in Section
III is investigated the ASE noise tolerance, in Section IV
is analyzed the crosstalk and optical filtering distortion,
Section V is devoted to chromatic dispersion, Section VI
measures the robustness of these formats to PMD, in
Section VII the nonlinear regime is analyzed, and finally
Section VIII concludes this paper.
2 Simulation scenario
The simulations have been performed with
computational-aided design tool OptiSystem. The
analysis of PDM-(D)QPSK depends largely each of the
devices that make up the optical network, so that the
results will be subject to the simulation scenario. For this
reason the simulations of the next sections have been
implemented over a generic model of DWDM network
(Fig. 2) [14]. The system includes 16 wavelengths (32
optical channels due to PDM). For 100 and 200-Gb/s
transmissions the spectral separation between super-
channels is 50 and 100 GHz respectively. The frequency
bands are 192.75 ‒ 193.5 THz in the first case and 192.4 ‒
193.9 THz in the second case. Transmission is performed
in 6x100-km spans with EDFA amplification. We have
Modulation Formats
100/200-Gbps
PDM-NRZ-(D)QPSK
PDM-67%RZ-(D)QPSK
PDM-50%RZ-(D)QPSK
PDM-33%RZ-(D)QPSK
PDM-CSRZ-(D)QPSK
Table 1 Different pulse shapes in PDM-(D)QPSK modulation.
Fig. 1. PDM-(D)QPSK modulation formats for 100 and 200-Gb/s
transmissions in DWDM systems.
employed standard single-mode fiber (SSMF) or non-zero
dispersion shifted fiber (NZDSF+) depending on the
simulation implemented on each section. The main fiber
parameters are shown in Table 8. We have included in the
optical link two multiplexers, two demultiplexers, a
reconfigurable optical add-drop multiplexer (ROADM)
and some optical filters to characterize the simulation
scenario as realistically as possible. In each optical
impairment discussed, the simulated scenario will have
slight variations, but all of them share some general
parameters (Table 2).
During the next sections the performance of PDM-
(D)QPSK will be analyzed in the presence of the main
physical impairments. The tolerance to ASE noise,
crosstalk, optical filtering, chromatic dispersion, PMD
and nonlinearities will be studied in order to discover the
pulse-shape (NRZ, RZ or CSRZ) which provides the best
performance in DWDM networks. QPSK and DQPSK
have the same temporal and spectral profile, so the
results of the analysis will be identical for both formats.
We will refer to these two modulations under the joint
notation of (D)QPSK. The only difference between them
is that QPSK carries the information into the phase of
optical pulses but DQPSK encodes the digital symbols
into the phase transitions.
3 ASE noise tolerance
Essential mechanisms of power losses in an optical fiber
are mainly derivatives of the absorption, spatial
dispersion and power radiated to the cladding [15]. In
long-haul transmissions, optical systems require optical
amplification because fiber losses reduce the signal power
below the detectability threshold of photodetectors.
Optical amplifiers can be designed as lumped elements
periodically spaced through the link forming several
amplification spans typically spaced 80 to 100 km in
terrestrial systems and 40 to 60 km in submarine
transmissions [2,15]. Optical amplification can also be
distributed by introducing gain along the transmission
fiber.
The main drawback of optical amplifiers in a
multispan scenario is the generation and accumulation of
amplified spontaneous emission (ASE) noise in the
optical spectrum. If multiple optical amplifiers are
concatenated to periodically compensate for fiber loss,
ASE builds up in the system, in analogy to the noise
build-up in an electrical amplifier chain. This noise build-
up is captured by the optical signal-to-noise ratio
(OSNR), which degrades with every amplifier along the
propagation path [16,17].
The ASE noise tolerance of 100 and 200-Gb/s PDM-
(D)QPSK signals is fully characterized by the required
OSNR (OSNRreq) [1], which is the OSNR that is needed
to achieve a specified target BER. Excluding FEC
overhead, the OSNRreq for a BERref = 10-12 was calculated
in a single-channel back-to-back (B2B) scheme, where
the transmitter is directly connected to the synchronous
homodyne coherent receiver [18,19], without filters or
optical fiber between TX and RX (Fig. 2). The goal is to
discover the pulse-shape (NRZ, RZ or CSRZ) that offers
better sensitivity. A higher sensitivity ensures greater
tolerance to degradation by accumulation of ASE noise in
the spectrum. Table 3 gives the results obtained in this
Parameter
Value
Laser linewidth (ECL)
100 KHz
MZM extinction ratio
30 dB
Noise Figure (EDFA)
5 dB
Dark Current (PIN)
10 nA
Responsivity (PIN)
0,9 A/W
Thermal noise
1x10-22 W/Hz
Bandwidth in optical filters
(Second-order Gaussian filters)
43 GHz (50 GHz grid)
85 GHz (100 GHz grid)
RF filters
Fourth-order Bessel filters
Cutoff frequency
0,75 x Rs
PRBS-Sequence length
32768 bits
Table 2 General parameters of the simulation scenario.
Fig. 2. Generic setup of a WDM system with polarization division multiplexing (PBC-Polarization Beam Combiner, PBS-Polarization
Beam Splitter).
Modulation Format
Coherent Detection
OSNRreq
(BERref=10-12)
100 Gb/s
OSNRreq
(BERref=10-12)
200 Gb/s
PDM-NRZ-(D)QPSK
21,0 dB
24,9 dB
PDM-67%RZ-(D)QPSK
20,3 dB
24,2 dB
PDM-50%RZ-(D)QPSK
20,2 dB
24,1 dB
PDM-33%RZ-(D)QPSK
19,8 dB
23,7 dB
PDM-CSRZ-(D)QPSK
20,2 dB
24,1 dB
Table 3 Sensitivity of PDM-(D)QPSK signals (0.1-nm resolution
bandwidth).
section. These values may differ from other works cited
in the bibliography due to various optical and electronic
hardware implementation aspects, including drive
waveforms, filter characteristics and modulator extinction
ratio [4,16,20]. Nevertheless, there are some general facts
which are worth mentioning.
QPSK and DQPSK, which are multilevel modulations,
require only 0.5-dB more OSNR than PSK and DPSK
signals in coherent and differential interferometric
detection [20,21]. Leaving aside TX/RX complexity
aspects, the good OSNR performance makes PDM-
(D)QPSK an attractive candidate for optically routed
networks that require a trade-off between sensitivity and
spectral efficiency. OSNR values listed in Table 3 haven
been measured in both polarizations and in a 12.5-GHz
optical reference bandwidth.
RZ pulses require ~1 dB less OSNR for identical BER
than NRZ. Particularly, signals with a low duty cycle
require less OSNR to obtain the specified target BER.
The shorter the pulse width, the higher the sensitivity of
PDM-(D)QPSK. We can see in Table 3 that a reduction in
the duty cycle decreases OSNRreq in the modulation
format. Consequently, PDM-33%RZ-(D)QPSK ends
emerging as the option that offers the best sensitivity. In
contrast, NRZ version has the lowest sensitivity with 21-
dB and 25-dB in 100 and 200-Gb/s transmissions,
respectively.
On the other hand, with PDM-CSRZ-(D)QPSK we
can achieve a sensitivity similar to PDM-50%RZ-
(D)QPSK with a higher duty cycle (67%). This is mostly
due to the reduced impact of ISI over the CSRZ pulses.
4 Crosstalk, optical filtering and spectral efficiency
Some formats are better suited than others when it comes
to tight WDM channel packing, quantified by its spectral
efficiency. Apart from important SE-dependent
nonlinearity considerations, in DWDM networks with
high spectral efficiency there are two concern
impairments arising from dense WDM channel spacing:
crosstalk and filter narrowing.
An optical channel can be affected by two different
types of crosstalk [19]: linear or interchannel crosstalk
(undesirable power of adjacent DWDM channels in the
desired band generating interferences at square-law
detection) and homodyne or intrachannel crosstalk (also
known as Multipath Interference or MPI [20], which
describes the coherent interference of a signal with
residual signals at the same wavelength due to imperfect,
reflective fiber connectors, double-Rayleigh
backscattering, or due to from imperfect drop capabilities
of OADMs…). The former is quite easy to avoid with
optimal filtering of the desired band, but the latter is
hardly removed by optical filters.
In general, phase modulated signals are more tolerant
to linear and homodyne crosstalk than intensity
modulations [21,22]. In analogy to linear crosstalk,
undesirable power in-band with the signal gives rise to
signal-MPI beat noise at square-law detection becoming
amplitude jitter in the eye diagram, so that the eye
penalty in PDM-(D)QPSK is lower than in OOK signals
due to the information is encoded in the optical phase.
Strictly, the impact of crosstalk on system performance
depends on the number of interferers, the OSNR
delivered, the modulation format and the signals
waveform (in particular the signal extinction ratio, and
the phase coherence of the interfering signals).
The tolerance of PDM-(D)QPSK to linear and
homodyne crosstalk has been analyzed. We have
compared different pulse shapes affected by the same
crosstalk ratio and we do not find major differences
between them. The narrowband PDM-NRZ-(D)QPSK
signal is less susceptible to generate linear crosstalk
penalties. Meanwhile, a broadband modulation spectrum
is less susceptible to MPI penalties due to reduced signal-
MPI beat noise (RZ formats) [20].
On the other hand, in order to narrow the bandwidth
of 100 and 200-Gb/s PDM-(D)QPSK signals, it is
necessary to filter them with the optimal bandpass to
avoid crosstalk impairments. However, the main problem
of optical filtering appears in DWDM networks with
high spectral efficiency, where the concatenation of
multiple ROADM’s narrows the overall optical filter
bandwidth and distorts the signals. In this situation, it is
particularly relevant to analyze the tolerance of PDM-
(D)QPSK to filtering distortion.
In order to simulate realistic conditions, the set of
network devices with filters in their architectures (mux,
demux, ROADM's ...) is modeled as an equivalent
cascade of filters between TX and RX in a B2B scenario.
We have analyzed two different cases:
a) 100-Gb/s in DWDM 50-GHz grid: five second-order
Gaussian bandpass filters with Full-Width at Half-
Maximum (FWHM) bandwidth of 43 GHz.
b) 200-Gb/s in DWDM 100-GHz grid: five second-order
Gaussian bandpass filters with FWHM bandwidth of
85 GHz.
The FWHM value is specified by ITU-T in the Rec.
G.694.1 (2012) for DWDM systems [10]. To quantify
filtering distortion we have measured the sensitivity of
these new schemes for a BERref = 10-12 and then we have
compared the results with the sensitivities of Table 3.
The values of this analysis are reported in Table 4.
In general, modulation formats with a narrower
bandwidth are more tolerant to optical filtering
distortion. Minimum distortion is shown in PDM-NRZ-
(D)QPSK signal with a 0.9-dB penalty in the OSNRreq.
The higher the pulse temporal width, the smaller the
spectral width of the signal, and therefore there will be a
lower penalty in the OSNR by optical filtering.
Accordingly, the tolerance to filtering distortion of PDM-
(D)QPSK is directly proportional to the value of the duty
cycle. These affirmations can easily be tested with the
results measured: if we increase the duty cycle value in
the optical carrier, filtering distortion decreases and the
OSNR penalty reaches its minimum value for the NRZ
pulses.
Particularly notable is the survey of PDM-33%RZ-
(D)QPSK. This signal has half of SE in DWDM systems
(1 b/s/Hz), making it impossible to use in filterless
optical networks [14]. Nevertheless, the distortion in its
waveform by the above filters is not excessively high.
This feature will prove to be essential. The handicap of
RZ pulses with a low duty cycle is an insufficient SE (<
2 b/s/Hz). However, prefiltering these signals in the TX,
the SE can be increased to 2 b/s/Hz (Table 5) with a
minimal penalty of ~1 dB in the OSNR (Table 4).
Figure 3 shows the eye diagram of PDM-33%RZ-
(D)QPSK. It is just mildly distorted at the output of the
filters chain. Filtering distortion results only in a slight
amplitude and phase jitter quantified in 1.3-dB OSNR
penalty. With 50% RZ and CSRZ PDM-(D)QPSK, the
spectral width of the main lobe ranges between 50% and
80% of bit rate, so these modulations can be considered
as narrowband signals for the above filters. Hence, the
OSNR penalty is so low (≈1 dB).
5 Chromatic dispersion (CD)
Chromatic dispersion is a major impairment in high-
capacity optical transmission systems. The spectral
components of the optical signal have different group
velocities, so that they reach the end of the fiber with
different group delays [23,24]. Consequently, chromatic
dispersion (also called Group Velocity Dispersion –
GVD) results in the time domain in dispersive pulse
broadening. Dispersion in optical fiber is an all-pass filter
on the electric field of the lightwave, given by a complex
Modulation Format
OSNRreq
(BERref=10-12)
Penalty OSNR
(Ref: table III)
100-Gb/s
200-Gb/s
PDM-NRZ-(D)QPSK
21,9 dB
25,8 dB
+0,9 dB
PDM-67%RZ-(D)QPSK
21,3 dB
25,2 dB
+1,0 dB
PDM-50%RZ-(D)QPSK
21,4 dB
25,3 dB
+1,2 dB
PDM-33%RZ-(D)QPSK
21,0 dB
25,0 dB
+1,3 dB
PDM-CSRZ-(D)QPSK
21,3 dB
25,2 dB
+1,1 dB
Table 4 Tolerance to optical filtering (0.1-nm resolution bandwidth).
Modulation Format
SEWDM
w/o filtering
(b/s/Hz)
SEWDM
with filtering
(b/s/Hz)
PDM-NRZ-(D)QPSK
2
2
PDM-67%RZ-(D)QPSK
1,87
2
PDM-50%RZ-(D)QPSK
1,25
2
PDM-33%RZ-(D)QPSK
1
2
PDM-CSRZ-(D)QPSK
1,54
2
Table 5 Bandwidth and SE of PDM-(D)QPSK in 100 and 200-Gb/s
DWDM transmissions.
Fig. 3. (a)-(b) Spectrum and eye diagram of PDM-33%RZ-(D)QPSK
without filtering. (c)-(d) Spectrum and eye diagram of PDM-33%RZ-
(D)QPSK with prefiltering. The distortion in the waveform of this signal by optical filtering is not excessively high, ~1 dB OSNR
penalty.
transfer function which shows a quadratic dependence in
its complex phase with the instantaneous frequency, like
a chirp filter. CD produces a variation of the
instantaneous frequency in a modulated optical signal
[25]. This variation of the instantaneous frequency
results in pulse broadening which generates inter-symbol
interference (ISI). Obviously, after propagating some
distance in the fiber, a point is reached where the
accumulating pulse spread is too great for the receiver to
recover the signal pulses within the equipment BER
specifications.
Table 6 quantifies the accumulated chromatic
dispersion (CDacum) required to induce a 2-dB penalty in
the OSNR at 200-Gb/s transmissions (100-Gb/s results
have been excluded due to space limitations). Obviously,
those signals that require more CDacum to reach the 2-dB
penalty show better tolerance to this impairment. The
second column presents the CD tolerance in filterless
optical networks [14], without optical filtering in the
simulation scenario (Fig. 2) and the third column
assumes the presence of filters. An optical filter disturbs
the waveform and the spectrum of the signals, so that the
tolerance to CD will also be affected. Currently, optical
networks have many devices with filtering in their
architectures so this analysis should be performed.
Graphs 4 and 5 also reflect the tolerance differences
to CD between PDM-(D)QPSK signals. We can see the
OSNR penalty evolution as a function of CDacum for each
signal, so that it is relatively simple to check out the
formats that are more tolerant to CD with and without
optical filtering in the simulation scenario. Attenuation,
PMD and fiber nonlinearities have been disabled in the
simulation scheme (Fig. 2) to measure the OSNR penalty
exclusively due to CDacum. Nonlinearities, like filtering,
may also modify the dispersion tolerance: SPM and
XPM broaden the bandwidth of the signals and hence
their ability to resist the CD. Nevertheless, intrachannel
nonlinearities prevail over interchannel nonlinear effects
in WDM systems beyond 40-Gbps/channel, so the results
listed in Table 6 are quite similar with and without Kerr
effect enabled. SSMF and NZDSF+ fibers have been
employed to measure the tolerance to CD as objectively
as possible (fiber parameters in Table 8). In addition, to
include the CD analysis in presence of optical filters in
the simulation scenario, five second-order Gaussian
Modulation Format
CDacum
w/o filtering
[ps/nm]
(2-dB pen.)
CDacum
FWHM = 85 GHz
[ps/nm]
(2-dB pen.)
PDM-NRZ-(D)QPSK
67,8
91,5
PDM-67%RZ-(D)QPSK
56,4
93,2
PDM-50%RZ-(D)QPSK
54,6
95,0
PDM-33%RZ-(D)QPSK
52,0
95,1
PDM-CSRZ-(D)QPSK
49,3
94,0
Table 6 Tolerance to chromatic dispersion for 200 Gb/s transmissions.
Fig. 4. OSNR penalty as a function of CDacum (ps/nm) without optical
filtering.
Fig. 5. OSNR penalty as a function of CDacum (ps/nm) with optical filtering.
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
OS
NR
Pen
alt
y (
dB
)
CDacum (ps/nm)
CD Tolerance (w/o filtering)
PDM-NRZ-(D)QPSK
PDM-33% RZ-(D)QPSK
PDM-50% RZ-(D)QPSK
PDM-67% RZ-(D)QPSK
PDM-CSRZ-(D)QPSK
1
1.5
2
2.5
3
3.5
4
4.5
5
70 80 90 100 110 120 130 140
OS
NR
Pen
alt
y (
dB
)
CDacum (ps/nm)
CD Tolerance (with filtering)
PDM-NRZ-(D)QPSK
PDM-33% RZ-(D)QPSK
PDM-50% RZ-(D)QPSK
PDM-67% RZ-(D)QPSK
PDM-CSRZ-(D)QPSK
bandpass filters were added with the same configuration
as in the previous section.
The tolerance of a modulation format to CD with and
without filtering is fixed by the spectrum and the
waveform launched to the fiber. In a simulation scenario
without optical filters these factors are exclusive of the
specific modulation employed in the lightpath. In
contrast, in the scenario with filtering the waveform and
the spectrum are determined by the set “modulation &
filters”. The narrower the signal spectrum, the higher the
tolerance to the CDacum. Filtering process reduces the
bandwidth of the signals, so the difference between the
group delays of the different spectral components will be
smaller.
In the simulation without filtering (the second column
of Table 6 and Fig. 4), the signals with more duty cycle
have a narrower spectrum, so PDM-NRZ-(D)QPSK and
PDM-67%RZ-(D)QPSK are the most tolerant to
chromatic dispersion. In contrast, 33% RZ and CSRZ
pulses present the worst performance. Particularly, PDM-
CSRZ-(D)QPSK, with the optical carrier suppressed, will
have better tolerance to fiber nonlinearities but this signal
loses robustness against accumulated chromatic
dispersion, as same as CSRZ-OOK format [20]. Despite
occupying a lower bandwidth than PDM-33%RZ-
(D)QPSK, the CSRZ version has the worst tolerance to
CD.
On the other hand, the order of the tolerance to CD
between the different versions is reversed with optical
filtering. As can be seen in the third column of Table 6
and Fig. 5, whereas PDM-NRZ-(D)QPSK shows more
resilience to CD than RZ versions in absence of filters,
PDM-RZ-(D)QPSK signals are more robust to CD in
presence of filters in the network. Filters guarantee
identical bandwidths in PDM-(D)QPSK formats, so that
CD tolerances are similar (the difference between the
group delays of the spectral components hardly vary
from one signal to another). Under identical conditions of
spectral width, the modulation with the narrowest
Gaussian pulses will show more robustness to CD. ISI
will be lower than in high duty cycle pulses for the same
CDacum value, so that PDM-33%RZ-(D)QPSK will have
the most resilience to CD in this case.
Prefiltering PDM-RZ-(D)QPSK formats it is possible
to increase their SE to 2 b/s/Hz in DWDM systems and
their tolerance to chromatic dispersion. Prefiltering
reduces the spectral components, so the difference
between the group delays is also reduced. Consequently,
the CD tolerance is increased. This feature is essential to
understand in Section 7 the great tolerance of this format
to intrachannel fiber Kerr nonlinearities, where the
chromatic dispersion plays a fundamental role.
6 Polarization mode dispersion
In ideal single-mode fibers, the two orthogonal modes
that compose the fundamental mode LP01 would be
degenerated with identical propagation properties
because they would have the same cutoff frequency and
the same propagation constant.
However, in real single-mode fibers, minute
waveguide asymmetries, either due to manufacturing
imperfections or due to stress imposed by mechanical
vibrations or temperature variations, the circular
symmetry between the core and cladding is not perfect
[26]. Under these conditions the two orthogonal modes
are non-degenerate. They have the same cutoff frequency
but their propagation constants are slightly different,
exhibiting different group velocities, and giving rise to a
Differential Group Delay (DGD) [15,27]. Because of the
difference between the group delays of both
polarizations, the initial pulse is broadened: After square-
law detection, the electrical signal is given by the
quadratic sum of both polarizations [20]:
𝑆(𝑡) = |𝐸𝑥(𝑡)|2 + |𝐸𝑦(𝑡 − 𝐷𝐺𝐷)|
2 (1)
This phenomenon is called Polarization Mode
Dispersion (PMD). If the DGD parameter is constant
over wavelength, it is referred to first-order PMD and it
predominates over superior PMD orders for many
applications. The DGD can be considered constant across
a single WDM channel, so the first-order PMD has been
studied in a single-channel transmission. The simulation
scenario is the same as Fig. 2 with only one TX/RX
(193.1 THz channel). A PMD emulator was added
between the TX and the RX. Table 7 and Figure 6 have
been obtained varying the DGD parameter in the PMD
emulator.
Table 7 quantifies the first order PMD tolerance of
PDM-(D)QPSK formats in 100 and 200-Gb/s
transmissions. It shows the DGD that leads to a 1-dB
Modulation Format
DGD-100Gb/s
(1-dB pen.)
DGD-200Gb/s
(1-dB pen.)
PDM-NRZ-(D)QPSK
19,0 ps
7,9 ps
PDM-67% RZ-(D)QPSK
19,4 ps
8,2 ps
PDM-50% RZ-(D)QPSK
19,6 ps
8,4 ps
PDM-33% RZ-(D)QPSK
20,0 ps
8,6 ps
PDM-CSRZ-(D)QPSK
19,6 ps
8,4 ps
Table 7 First-order PMD tolerance. It shows the DGD (ps) that leads
to a 1-dB OSNR penalty.
OSNR penalty. The first-order PMD tolerance of a
modulation format is linear with the symbol period [28].
For most of them, a 1-dB penalty occurs at a DGD value
between 30% and 40% of the symbol period. The
tolerance of PDM-(D)QPSK formats is very similar and
there are no major differences between them, with RZ
formats being in general more resilient to PMD than the
NRZ version.
The worst result is observed using the NRZ pulses.
The Non-Return to Zero pulses, occupying the entire
symbol interval, overlap each other with a lower DGD
between polarizations. In contrast, PDM-33%RZ-
(D)QPSK is the most resilient to PMD (DGDMAX). The
reason is simple: the narrower the RZ pulses, the higher
DGD required to overlap adjacent pulses with the same
ISI-penalty. Therefore, PDM-33%RZ-(D)QPSK requires
a higher DGD between polarizations to suffer the same
OSNR penalty as pulses with more duty cycle.
Meanwhile, PDM-CSRZ-(D)QPSK shows the same
tolerance to first-order PMD as a 50% RZ despite having
a 67% duty cycle value.
Additionally, Fig. 6 shows the OSNR penalty as a
function of DGD for 200G-PDM-(D)QPSK signals (100-
Gb/s results are not included due to space limitations, but
the order of the tolerances to PMD between 100G-PDM-
(D)QPSK formats remains unalterable). It is also obvious
that the signal with the lowest duty cycle is the most
robust to PMD.
On the other hand, note that the resilience to PMD, in
addition, depends to an appreciable extent on the
waveforms and filters as well as other residual distortions
such as CD and fiber Kerr nonlinearities [19].
7 Fiber Kerr nonlinearities
The effective area of a single-mode fiber typically ranges
between 20 and 100 μm2 resulting in a strong
confinement of the light within the fiber core, so that the
optical intensity can easily exceed MW/cm2. Due to the
presence of an optical intensity so high, the refractive
index value could be affected resulting in a phase
deviation of the fundamental mode. The refractive index
dependence with the optical intensity is known as the
Kerr effect [29].
In order to better understand the origin of
nonlinearities, the Kerr effect can be classified into more
specific nonlinear interactions that can distort optical
signals in different ways (Fig. 7). Fiber nonlinearities
occurring between pulses of the same WDM channel are
referred to as intrachannel nonlinearities and the
nonlinearities occurring among two or more WDM
channels, the expression interchannel nonlinearities is
used [28].
The nonlinear regime of a single-polarization
electrical field through a single-mode optical fiber is
described by the Generalized Nonlinear Schrödinger
Equation (GNLSE) [20,24,29]:
𝜕�⃗� 1(𝑧, 𝑡)
𝜕𝑧+𝑗
2𝛽2(𝑧)
𝜕2�⃗� 1(𝑧, 𝑡)
𝜕𝑡2−1
6𝛽3(𝑧)
𝜕3�⃗� 1(𝑧, 𝑡)
𝜕𝑡3+
+𝛼(𝑧)
2�⃗� 1(𝑧, 𝑡) = 𝑗𝛾|�⃗� 1|
2�⃗� 1⏟
𝑺𝑷𝑴−𝑰𝑺𝑷𝑴
+ 2𝑗𝛾 {|�⃗� 2|2+ |�⃗� 3|
2} �⃗� 1⏟
𝑿𝑷𝑴−𝑰𝑿𝑷𝑴
+
+ 𝑗𝛾�⃗� 1�⃗� 2�⃗� 3∗ ⏟
𝑭𝑾𝑴−𝑰𝑭𝑾𝑴
(2)
In intrachannel nonlinearities E⃗⃗ 1, E⃗⃗ 2 and E⃗⃗ 3 represent
the electric fields associated to three different optical
pulses from the same WDM channel, meanwhile they
represent three different WDM channels in interchannel
nonlinearities.
The nonlinear interaction of a channel or a pulse with
itself is referred to as self-phase modulation (SPM).
Whether SPM relates to an entire channel or an isolated
pulse (ISPM) depends on the context [19]. ISPM
explains the phase variations in E⃗⃗ 1 which are generated
by fluctuations in its own intensity, considering the
isolated pulse and without the intervention of adjacent
pulses [20]. The intensity variations in E⃗⃗ 1 cause changes
Fig. 6. OSNR penalty as a function of DGD [ps].
0.5
1
1.5
2
2.5
3
6 7 8 9 10 11 12 13 14 15 16 17 18
OS
NR
Pen
alt
y (
dB
)
DGD (ps)
PMD Tolerance
PDM-NRZ-(D)QPSK
PDM-33% RZ-(D)QPSK
PDM-50% RZ-(D)QPSK
PDM-67% RZ-(D)QPSK
PDM-CSRZ-(D)QPSK
in the refractive index resulting in variations in its own
phase.
Intrachannel cross-phase modulation (IXPM) and
intrachannel four-wave mixing (IFWM) are the others
intrachannel nonlinearities arising from the signal-signal
interactions. They are the dominant nonlinear effects
beyond 40-Gb/s per channel transmissions. Both
phenomena are generated as a result of the overlapping
between pulses in a single optical carrier due to CD at the
same time that the Kerr effect is being stimulated
[19,30]. IXPM explains the pulse phase variations due to
intensity fluctuations in the overlapped adjacent pulses.
On the other hand, IFWM explains the interaction
between three different pulses within the same optical
channel. The pulses interaction generates a fourth pulse
with random amplitude, often referred as ghost pulse
[19,28]. The IXPM effects can be visualized in the eye
diagram as phase jitter while IFWM generates amplitude
jitter. In Fig. 8 both concepts can be visualized.
The intensity fluctuations generated by ASE noise
and adhered over the pulse amplitude are potentially
hazardous in presence of SPM. They induce random
phase fluctuations (signal-ASE interactions). This phase
jitter is a new noise known as nonlinear phase noise
(NPN), also referred as the Gordon-Mollenauer effect
[17,19,28,31] and it is particularly detrimental for phase-
modulated systems between 1 and 20-Gb/s. Beyond 40-
Gb/s, the Gordon-Mollenauer effect is not an important
nonlinear impairment [21], although optical signals are
affected by NPN from intrachannel nonlinearities due to
signal-ASE interactions. NPN is especially detrimental in
dispersion managed (DM) systems with phase-modulated
signals.
Two new phenomena appear associated with signal-
signal interactions in interchannel nonlinearities: cross-
phase modulation (XPM) and four-wave mixing (FWM)
[29]. These nonlinear effects are assumed widely known
so they are not described in this paper.
The easiest way to reduce the XPM effects is
increasing the spectral separation between optical
carriers or using optical fibers with a higher chromatic
dispersion coefficient to avoid the spatial overlap
between pulses of different WDM channels [15,32].
Beyond 20-Gb/s, XPM is not an important impairment in
nonlinear regime [28,29]. Considering the signal-ASE
interaction, XPM can induce NPN in adjacent channels.
In contrast with the Gordon-Mollenauer effect, NPN
induced by XPM cannot be compensated because it is not
correlated with the received optical intensity [19],
although its impact in the OSNR is lower than the
Gordon-Mollenauer effect in SSMF fibers in 10-Gb/s
systems [21]. A major consequence of SPM and XPM
(signal-signal and signal-ASE interactions) is the spectral
broadening of the pulses that increases the channels
bandwidth considerably and limits the performance of a
lightwave system [25]. Therefore, both phenomena
contribute to reduce the system tolerance to chromatic
dispersion [28].
On the other hand, FWM is generated, like XPM, due
to the nonlinear response of the dielectric polarization
with the electric field of the light. Nevertheless, FWM
will cease to be a source of degradation in the nonlinear
regime beyond 80-Gb/s transmissions [21,32]. According
to Winzer’s studies [20,21,28,31], intrachannel
nonlinearities only prevail beyond 40-Gb/s per channel
transmissions: IFWM in SSMF spans and IXPM in
NZDSF+ spans. Beyond 200-Gb/s, IFWM is the major
impairment in the nonlinear regime.
Fig. 7. Classification of Kerr nonlinearities.
Fig. 8. Effects of IXPM and IFWM on a 100 and 200-Gb/s PDM-
(D)QPSK signals. The upper graphs show the signal waveform after
transmission. The lower graphs show the eye diagram. The main effect
of IXPM is to produce timing jitter, meanwhile IFWM induces amplitude jitter (ghost pulses).
7.1 IXPM and IFWM tolerance
The simulations of this section have been designed to
survey the tolerance of 100/200G-PDM-(D)QPSK
signals to IXPM and IFWM, the dominant nonlinearities
beyond 40-Gb/s. In this section, the simulations have
been supported on the DWDM system shown in Fig. 2.
The analysis in the nonlinear regime depends on many
parameters of the simulation scenario and its operating
conditions [33], so one should be very careful with the
measurements to obtain conclusions as objective as
possible. The presence of IXPM and IFWM is directly
related to the GVD coefficient of the fiber, so both
phenomena have been analyzed for SSMF and NZDSF+
spans (parameters in Table 8).
Firstly, we initially tested the absence of FWM and
XPM in 100 and 200-Gb/s transmissions. DWDM
spectrum was controlled along the lightpath, checking
that no new spectral components were generated (FWM
not found) and the bandwidth of each channel remained
unchanged (XPM not found) during the propagation.
These tests ensured that the predominant nonlinearities
were IXPM and IFWM for both types of fibers.
Afterwards, in Fig. 9, we calculated the OSNR
penalty evolution as a function of the power launched
into the fiber to find the pulse-shape that is the most
robust to IXPM and IFWM in PDM-(D)QPSK signals.
We only show the graph for the 200-Gbps-SSMF case
because the tolerance order to nonlinear regime among
the different pulses does not vary in the other analyzed
cases (NZDSF+ and 100-Gb/s). Although the graph for
the NZDSF+ has not been included due to space
limitations in this document, it is interesting to compare
with the SSMF in the nonlinear regime. The robustness
to nonlinearities can be increased using the standard
single-mode fiber [36] due to a higher CD coefficient, a
higher effective area and a lower nonlinear coefficient.
Attending to its CD coefficient, optical pulses will be
more broadened with a lower distance than in the
NZDSF+, so that the peak power of the temporal profile
of the pulses will drop before and they will propagate
fewer kilometers stimulating the Kerr effect [37]. For the
same value of Plaunch (dBm), the Kerr effect will be
stimulated in the SSMF spans during a shorter distance
than in NZDSF+ spans.
PDM-(D)QPSK signals exhibit an excellent tolerance
to intrachannel nonlinearities. In 100 and 200-Gb/s
transmissions, IFWM appears in the SSMF and NZDSF+
whereas IXPM is mainly stimulated in the NZDSF+. In
general, phase modulated signals are more robust to
IFWM than OOK formats because the relative phase
between pulses does not vanish and hence the ghost
pulses generated by IFWM have a lower amplitude.
33%-RZ and CSRZ pulses are the most tolerant to IFWM
and IXPM. Knowing that IFWM and IXPM arise from
pulse-to-pulse interactions, the shorter the pulse width,
the higher the tolerance to these phenomena [30,38].
With a low duty cycle, RZ pulses need more CDacum to
overlap with each other. In that case, there will be a
lower peak power in optical pulses with more CDacum, so
the Kerr effect will be less stimulated. Therefore, a
higher optical power will be necessary to launch into the
fiber to achieve the same OSNR penalty as in pulses with
a higher duty cycle. Accordingly to Fig. 9, the worst
tolerance to IFWM and IXPM is found in the PDM-
NRZ-(D)QPSK signal.
However, the high robustness offered by PDM-
CSRZ-(D)QPSK is nearly identical to the 33% RZ
version. A large resistance to nonlinearities is achieved
Optical Fiber
SSMF
NZDSF+
D [ps/(nm∙km)]
+17
+4
S [ps/(nm2∙km)]
+0.088
+0.084
ATH [dB]
0.2
0.2
Aeff [μm2]
80
72
n2 x10-20 [m2/W]
2.60
3.53
DPMD [ps/km1/2]
< 0.1
< 0.1
Table 8 Parameters of SSMF and NZDSF+ fibers [34,35].
Fig. 9. Tolerance to IXPM and IFWM. OSNR penalty as a function of
Plaunch (dBm).
0
1
2
3
4
5
4 5 6 7 8 9 10 11 12 13 14 15 16 17
OS
NR
Pen
alt
y (
dB
)
Plaunch (dBm)
IXPM and IFWM tolerance (200-Gb/s)
PDM-NRZ-(D)QPSK
PDM-33% RZ-(D)QPSK
PDM-50% RZ-(D)QPSK
PDM-67% RZ-(D)QPSK
PDM-CSRZ-(D)QPSK
suppressing the optical carrier in RZ pulses despite
having a higher duty cycle (67%).
7.2 Variation of IXPM and IFWM with the duty cycle
In 100 and 200-Gb/s transmissions, PDM-(D)QPSK
signals carry 50 and 100-Gb/s per polarization,
respectively (Fig. 1). According to Peter J. Winzer in
[28], IFWM predominates in SSMF spans whereas
IXPM appears in NZDSF+ spans with bit rates ranging
between 50 and 100-Gb/s per polarization. However, the
presence of both nonlinearities does not only depend on
the optical fiber type, but also the duty cycle value must
be taken into account.
A priori, IFWM prevails over IXPM in high
dispersion fibers. This is mainly valid with low duty
cycles (33% ‒ 50%), but if we set up the optical carrier
with a higher duty cycle value (67% ‒ 100%) the
stimulation of IXPM will increase, i.e. IXPM is
proportional to the duty cycle value. Obviously, the pulse
overlapping is more pronounced with a higher duty cycle
and consequently the phase is more distorted by the
amplitude fluctuations of adjacent pulses. In this
situation, the eye diagram is closed by amplitude jitter
(associated with the presence of ghost pulses generated
by IFWM) and phase jitter (due to NPN generated by
IXPM). Therefore, IXPM cannot be ignored in SSMF
spans for the PDM-NRZ-(D)QPSK and PDM-67%RZ-
(D)QPSK signals (Fig. 10.a).
On the other hand, the differences are not so
pronounced in low dispersion fibers for different duty
cycle values (Fig. 10.b). In these fibers, PDM-(D)QPSK
signals are affected by both intrachannel nonlinearities,
without a clear predominance of any of them. For low
duty cycles, IFWM is slightly lower than IXPM, so the
phase jitter is the main cause of the eye diagram
degradation. Meanwhile, IFWM increases with the duty
cycle value and puts on the same level as IXPM. In that
case, the eye diagram is affected by amplitude and phase
jitter.
In general, a high duty cycle value favors the
presence of both intrachannel nonlinearities, whereas the
GVD coefficient is fundamental to determine the
predominant nonlinearity when a low duty cycle is
employed: IFWM in SSMF and IXPM in NZDSF+.
8 Conclusions
The amount of traffic carried on optical backbone
networks has been growing exponentially over the past
two decades. Nowadays, the required capacity on each
DWDM channel ranges between 100 and 200-Gb/s. In
2012, the ITU-T recommended the use of “Flexible-
Grids” networks enabling the possibility to continue
working with the PDM-(D)QPSK format beyond 100-G,
due to the current technological difficulties to achieve
long-haul data communications with the multilevel
PDM-16-QAM format, mainly focused on short-reach
transmissions.
An intensive analysis has been made of polarization-
division multiplexed quadrature phase shift keying
modulation formats in 100 and 200-Gb/s DWDM
systems. The performance offered by PDM-(D)QPSK
varies according to the line code that is employed to
carve the pulses of the optical carrier: NRZ, 33% RZ,
50% RZ, 67% RZ or CSRZ. Each option offers different
tolerances to linear and nonlinear impairments of a
lightpath. The features of PDM-(D)QPSK have been
characterized measuring the robustness to ASE noise,
Fig. 10. Eye diagrams of PDM-33%RZ-(D)QPSK (left) and PDM-NRZ-(D)QPSK (right). The distortion induced by IXPM and IFWM
results in timing and amplitude jitter, respectively.
crosstalk, optical filtering, chromatic dispersion,
polarization mode dispersion and intrachannel
nonlinearities.
PDM-33%RZ-(D)QPSK reveals an enormous
resistance against the ghost pulses generated by IFWM
and the phase fluctuations of IXPM, but the handicap of
this signal is its low SE in DWDM systems, 1 b/s/Hz.
Thanks to its great tolerance to filtering distortion, the SE
may be increased from 1 to 2 b/s/Hz in DWDM systems
prefiltering this format in the TX and allowing its
integration in the grids of 50 and 100 GHz for 100 and
200 Gb/s transmissions, respectively. In this way, it is
possible to exploit the great advantage of RZ pulses with
low duty cycle: the reduction of pulse-to-pulse
interaction. This feature allows to increase the tolerance
to CD, PMD and intrachannel nonlinearities. In contrast,
the PDM-NRZ-(D)QPSK format has a slightly higher SE
than the other options and it shows better tolerance to
filtering distortion and chromatic dispersion in filterless
networks.
References
[1] Y. Ayhan, I.H. Cavdar, Optimum link distance determination
for a constant signal to noise ratio in M-ary PSK modulated
coherent optical OFDM systems, Telecommunication
Systems 55 (2014) 461-470.
[2] M. Salsi, J. Renaudier, O. Bertran-Pardo, H. Mardoyan, P.
Tran, G. Charlet, S. Bigo, 100 Gb/s and Beyond for
Submarine Systems, J. Lightw. Technol. 30 (2012), 3880-
3887.
[3] P.J. Winzer, High-Spectral-Efficiency Optical Modulation
Formats, J. Lightw. Technol. 30 (2012), 3824-3835.
[4] L. Li, Z. Jijun, D. Degong, Y. Aihan, Analysis modulation
formats of DQPSK in WDM-PON system, Optik 123 (2012),
2050-2055.
[5] G. Bennett, Superchannels to the rescue, Lightwave,
March/April 2012.
[6] R. Saunders, Coherent DWDM technology for high speed
optical communications, Optical Fiber Technology 17
(2011), 445–451.
[7] S. Matthias, Systems with Higher-Order Modulation Impact
of Nonlinearities on Fiber Optic Communications”, Optical
and Fiber Communications Reports 7 (2011), 177-217.
[8] R.H. Walden, Analog-to-digital converters and associated IC
technologies, Proc. Compound Semiconductor Integrated
Circuits Symposium (2008), 1-2.
[9] J. Yu, Z. Dong, H.-C. Chien, Z. Jia, X. Li, D. Huo, M.
Gunkel, P. Wagner, H. Mayer, A. Schippel, Transmission of
200 G PDM-CSRZ-QPSK and PDM-16QAM With a SE of 4
b/s/Hz, J. Lightw. Technol. 31 (2013), 515-522.
[10] ITU-T G.694.1 Recommendation (2012), Spectral grids for
WDM applications: DWDM frequency grid.
[11] J. Wang, C. Xie, Z. Pan, Generation of Spectrally Efficient
Nyquist-WDM QPSK Signals Using Digital FIR or FDE
Filters at Transmitter, J. Lightw. Technol. 30 (2012), 3679-
3686.
[12] G. Bosco, P. Poggiolini, M. Visintin, Performance Analysis
of MLSE Receivers Based on the Square-Root Metric, J.
Lightw. Technol. 26 (2008) 2098-2109.
[13] J. Tellado, M. Louise, C. Hoo, J.M. Cioffi, Maximum-
Likelihood Detection of Nonlinearly Distorted Multicarrier
Symbols by Iterative Decoding, IEEE Transactions on
Communications 51 (2003), 218-228.
[14] C. Tremblay, É. Archambault, M.P. Bélanger, J.-P. Savoie,
F. Gagnon, D.V. Plant, Passive filterless core networks based
on advanced modulation and electrical compensation
technologies, Telecommunication Systems 54 (2013), 167-
181.
[15] J.A.M. Pereda, Sistemas y Redes Ópticas de
Comunicaciones, Pearson Education, 2004.
[16] A.H. Gnauck, Advanced Amplitude and Phase Coded
Formats for 40Gb/s Fiber Transmission, Proceedings of
IEEE/LEOS Annual Meeting, page WR1, 2004.
[17] J.P. Gordon, L.F. Mollenauer, Phase noise in photonic
communications systems using linear amplifiers, Optics
Letters 15 (1990), 1351–1353.
[18] G. Li, Recent Advances in Coherent Optical
Communications, Advances in Optics and Photonics 1
(2009), 279-307.
[19] K.-P. Ho, Phase-Modulated Optical Communications
Systems, Springer, 2005.
[20] P.J. Winzer, R.J Essiambre, Advanced optical modulation
formats, J. Lightw. Technol. 24 (2006), 4711-4727.
[21] P.J. Winzer, A.H. Gnauck, Optical Phase-Shift-Keyed
Transmision, J. Lightw. Technol. 23(2005), 115–130.
[22] Y. Aihan, L. Li, Z. Xinliang, Analysis of modulation format
in the 40Gbit/s optical communication system, Optik 121,
(2010) 1550–1557.
[23] G.P. Agrawal, Fiber-Optic Communication Systems, Wiley
Interscience, Four Edition, 2010.
[24] J.M. Senior, Optical Fiber Communications, Principles and
Practice, Prentice Hall. 2009.
[25] B.A.E. Saleh, M.C. Teich, Fundamentals of Photonics, John
Wiley & Sons, 1991.
[26] J.P. Gordon, PMD fundamentals: Polarization mode
dispersion in optical fibers, PNAS 97 (2000), 4541-4550.
[27] J.P. Elbers, Modelling of polarization mode dispersion in
single mode fibers, Electronics Letters 33 (1997), 1894-1895.
[28] P.J. Winzer, R.J Essiambre, Advanced optical modulation
formats, Optical Fiber Telecommunications V B: Systems
and Networks, Elsevier, 2008.
[29] G.P. Agrawal, Nonlinear Fiber Optics, San Diego: Elsevier
Science & Technology, 5th edition, 2013.
[30] L. Lujiao, Q. Yaojun, J. Yuefeng, Suppression of intra-
channel four-wave mixing in 40 Gbit/s RZ-DQPSK
transmission with alternate-polarization, Optik 122 (2011),
2242– 2245. [31] H. Kim, P.J. Winzer, Nonlinear Phase Noise in Phase-Coded
Transmission, in Proc. OFC, page OThO3, 2005.
[32] D.P. Abellán, F.P. Ramos, J.C. Francoy, Sistemas de
Comunicaciones Ópticas, Universidad Politécnica de
Valencia, 2006.
[33] R. Hui, M. O’Sullivan, Fiber Optic Measurement
Techniques, Elsevier Academic Express, 2009.
[34] Corning Incorporated, Datasheet Corning® LEAF® Optical
Fiber, 2011.
[35] Corning Incorporated, Datasheet Corning® SMF-28e+®
Optical Fiber, 2011.
[36] M.S. Alfiad, D. Van den Borne, T. Wuth, M. Kuschnerov, H.
Waardt, On the Tolerance of 111-Gb/s POLMUX-RZ-
DQPSK to Nonlinear Transmission Effects”, J. Lightw.
Technol. 29 (2011) 162-170.
[37] M.Y. Hamza, N. Akhtar, N. Sarwar, S. Yang, Evolution
behavior of chirped tan-hyperbolic pulse through single
mode fiber in the simultaneous presence of fiber loss,
dispersion and self-phase modulation, Telecommunication
Systems 55 (2014), 451-459.
[38] K.S. Cheng, J. Conradi, Reduction of pulse-to-pulse
interaction using alternative RZ formats in 40-Gb/s systems,
IEEE Photon. Technol. Lett. 14 (2002), 98–100.